O. Berteau, INRA, UMR1319 MICALIS, Bât 440, Domaine de Vilvert, F-78352 Jouy-en-Josas, France Fax: +33 1346 52462 Tel: +33 1346 52308 E-mail: email@example.com
Sulfatases form a major group of enzymes present in prokaryotes and eukaryotes. This class of hydrolases is unique in requiring essential post-translational modification of a critical active-site cysteinyl or seryl residue to Cα-formylglycine (FGly). Herein, we report mechanistic investigations of a unique class of radical-S-adenosyl-L-methionine (AdoMet) enzymes, namely anaerobic sulfatase-maturating enzymes (anSMEs), which catalyze the oxidation of Cys-type and Ser-type sulfatases and possess three [4Fe-4S]2+,+ clusters. We were able to develop a reliable quantitative enzymatic assay that allowed the direct measurement of FGly production and AdoMet cleavage. The results demonstrate stoichiometric coupling of AdoMet cleavage and FGly formation using peptide substrates with cysteinyl or seryl active-site residues. Analytical and EPR studies of the reconstituted wild-type enzyme and cysteinyl cluster mutants indicate the presence of three almost isopotential [4Fe-4S]2+,+ clusters, each of which is required for the generation of FGly in vitro. More surprisingly, our data indicate that the two additional [4Fe-4S]2+,+ clusters are required to obtain efficient reductive cleavage of AdoMet, suggesting their involvement in the reduction of the radical AdoMet [4Fe-4S]2+,+ center. These results, in addition to the recent demonstration of direct abstraction by anSMEs of the Cβ H-atom from the sulfatase active-site cysteinyl or seryl residue using a 5′-deoxyadenosyl radical, provide new insights into the mechanism of this new class of radical-AdoMet enzymes.
Sulfatases belong to at least three mechanistically distinct groups, namely the Fe(II) α-ketoglutarate-dependent dioxygenases , the recently identified group of Zn-dependent alkylsulfatases  and the broad family of arylsulfatases . This latter family of enzymes, termed ‘sulfatases’ in this article, is certainly the most widespread among bacteria, some of which possess more than 100 sulfatase genes in their genomes . Nevertheless, their biological function has almost never been investigated despite reports on their potential involvement in pathogenic processes [5,6].
Among hydrolases, sulfatases are unique in requiring an essential catalytic residue, a 3-oxoalanine, usually called Cα-formylglycine (FGly) . In sulfatases, it has been proposed that this modified amino acid is hydrated as a geminal diol in order to perform a nucleophilic attack on the sulfur atom of the substrate. This leads to the release of the desulfated product and the formation of a covalent sulfate–enzyme intermediate. The second hydroxyl group of the germinal diol is essential for the release of the inorganic sulfate, as demonstrated by the inactivation of a sulfatase bearing a seryl residue instead of the FGly residue .
This essential FGly residue results from the post-translational modification of a critical active-site cysteinyl or seryl residue (Fig. 1A). This has led to the classification of sulfatases into two subtypes, namely Cys-type sulfatases and Ser-type sulfatases. In eukaryotes, only Cys-type sulfatases have been identified so far, while in bacteria, both types of sulfatases exist. Nevertheless, eukaryotic and prokaryotic sulfatases undergo identical post-translational modification involving the oxidation of a critical cysteinyl or a seryl residue into 3-oxoalanine.
In prokaryotes, 3-oxoalanine formation is catalyzed by at least three enzymatic systems but to date only two have been identified . The first enzymatic system, termed formylglycine-generating enzyme, uses molecular oxygen and an unidentified reducing agent to catalyze the aerobic conversion of the cysteinyl residue into FGly . The second enzymatic system, termed anaerobic sulfatase maturating enzyme (anSME), is a member of the S-adenosyl-L-methionine (AdoMet)-dependent superfamily of radical enzymes [11–13].
We have recently demonstrated that anSMEs are dual-substrate enzymes with the ability to catalyze the oxidation of cysteinyl or seryl residues, making these enzymes responsible for the activation of both types of sulfatase under anaerobic conditions . Nevertheless, the mechanism by which these enzymes catalyze the anaerobic oxidation of cysteinyl or seryl residues is still obscure. Furthermore, in addition to the Cx3Cx2C motif that binds the [4Fe-4S]2+,+ cluster common to all radical AdoMet superfamily enzymes, anSMEs have two additional conserved cysteinyl clusters with unknown functions.
In the present study, we carried out mutagenesis studies to investigate the involvement of the conserved cysteinyl clusters in the anSME’s mechanism. Our data demonstrate that the additional conserved cysteinyl clusters bind two additional [4Fe-4S ]2+,+ centers that are required for the generation of FGly and for the efficient reductive cleavage of AdoMet, suggesting that one or both of the additional [4Fe-4S]2+,+ centers play a role in mediating the reduction of the radical-AdoMet [4Fe-4S]2+,+ cluster.
Formylglycine and 5′-deoxyadenosine kinetics
The first step of the reaction catalyzed by all radical AdoMet enzymes investigated thus far is the reductive cleavage of AdoMet, via one-electron transfer from the enzyme [4Fe-4S]+ center to AdoMet, to yield methionine and a 5′-deoxyadenosyl radical [14,15]. AdoMet is generally used as an oxidizing substrate, with the notable exception of enzymes such as lysine 2,3-aminomutase [15,16] and spore photoproduct lyase [17–20], which use AdoMet catalytically. In other radical AdoMet enzymes, AdoMet is a co-substrate and, as such, one equivalent of AdoMet is used to oxidize one molecule of substrate. The only known exceptions are coproporphyrinogen III oxidase (HemN), which uses two AdoMet molecules per turnover for the decarboxylation of two propioniate side chains [21,22], and the radical AdoMet enzymes, which catalyze sulfur insertion, such as lipoyl synthase, biotin synthase and MiaB [14,15].
Recently, Grove et al. characterized the Klebsiella pneumoniae anSME (anSMEkp) and investigated the maturation of a 18-mer peptide, derived from the K. pneumoniae sulfatase sequence, containing the seryl residue target of the modification . Quantitative data were extracted from HPLC and MALDI-TOF MS analyses of the products. With the 18-mer peptide substrate, three uncharacterized products and 5′-deoxyadenosine (5′-dA) were observed using HPLC analysis, and two peptide products were identified using MS analysis. The expected FGly product (i.e. a 2 Da mass decrease, see Fig. 1A) was found to be a minor product in the MS analysis, while the major product exhibited a 20 Da mass decrease, which was tentatively attributed to the loss of a water molecule from the FGly product as a result of the formation of a Schiff base via an interaction between the aldehyde carbonyl of FGly and the N-terminal amino group. The three products observed in the HPLC analysis were not further characterized and it is not currently possible to state whether or not they are FGly-containing peptides, reaction by-products or reaction intermediates. Nevertheless, based on the assumption that all three products observed by HPLC corresponded to, or were derived from, the FGly product, the authors concluded that anSMEs use one mole of AdoMet to produce one mole of FGly-containing peptide. While this is the most likely scenario based on mechanistic studies of other radical AdoMet enzymes, this result must be viewed as preliminary in light of the undetermined nature of the multiple peptide products.
Intrigued by the possibility that some of the peptides produced could be reaction intermediates, we performed similar experiments with the Clostridium perfringens anSME (anSMEcpe) that was recently characterized in our laboratory [11,12]. In our previous studies, we used 23-mer peptides as substrates [11,12]. Although these substrates proved to be satisfactory to demonstrate that anSMEs are able to catalyze the anaerobic oxidation of cysteinyl or seryl residues, the instability of these peptides prevented accurate quantifications of the enzymatic reaction. We thus investigated several peptides in order to identify a more stable substrate and finally chose a 17-mer peptide, which is closer in size to the 18-mer substrates used by Grove et al. . The substrate peptides used were Ac-TAVPSCIPSRASILTGM-NH2 (17C peptide) ([M+H]+ = 1745) and Ac-TAVPSSIPSRASILTGM-NH2 (17S peptide) ([M+H]+ = 1729). Upon incubation with anSMEcpe, both peptides were converted into a new species with a mass [M+H]+ of 1727 Da (Figs1B,C and S1). This molecular mass was precisely the one expected for the conversion of the cysteinyl residue or the seryl residue into FGly. To further ascertain the nature of the modification, labeling experiments with 2,4-dinitrophenyl-hydrazine (DNPH) were performed . A hydrazone derivative with a mass increment of 180 Da was formed, demonstrating the presence of an aldehyde functional group in the newly formed peptide (Fig. S2). Thus, in our experiments, only the substrate and the expected product were evident in the mass spectra and no other species appeared, even after extended incubation (i.e. 12 h with peptide 17S) (Figs 1, S1 and S2).
We then developed an HPLC-based assay that could provide reliable and direct quantitative data regarding the anSME activity. During incubation with each peptide, one new peptide appeared with a retention time of 20.4 min (Fig. 2A,B). The purification of this product and its MALDI-TOF MS analysis confirmed the nature of the product formed, and kinetic experiments demonstrated that, in both cases (i.e. with a cysteinyl-containing peptide or a seryl-containing peptide) a strict 1 : 1 coupling between AdoMet cleavage and FGly production occurred (Fig. 2C,D). AnSMEcpe exhibited a specific activity of 0.07 nmol·min−1·mg−1 with the 17S substrate, whereas the specific activity increased by more than 15-fold (to 1.09 nmol·min−1·mg−1) for the 17C substrate.
Peptide 17A was initially included as a control to demonstrate that FGly production occurred on the target cysteinyl or seryl residue. As expected, in the presence of enzyme, no modification of the peptide 17A occurred (Figs 2 and S1C). Interestingly, AdoMet cleavage analysis in the presence of peptide 17A showed that no 5′-dA was produced (Fig. 2D). This result is surprising because we previously showed that anSMEcpe, alone, is able, under reducing conditions using sodium dithionite as electron donor, to produce 5′-dA from AdoMet . This result suggests that nonproductive peptides, such as 17A, bind near the active site and prevent either direct reduction of the [4Fe-4S]2+,+ center or interaction with new AdoMet molecules.
Analytical and spectroscopic evidence for multiple Fe-S clusters in anSME
We previously demonstrated that anSMEs possess a typical radical AdoMet [4Fe-4S]2+,+ center that is probably coordinated, as in all radical AdoMet enzymes, by the Cx3Cx2C motif . Interestingly, in addition to this first conserved cysteine motif, anSMEs have seven other strictly conserved cysteinyl residues and an additional cysteinyl residue in the C-terminus part of the protein (Fig. 3A). We and other groups [11,12,25,26] have proposed that additional iron–sulfur cluster(s) may be coordinated by the remaining conserved cysteinyl residues. Nevertheless, in our previous analytical and spectroscopic studies of as-purified and reconstituted samples of wild-type (WT) anSMEcpe, we did not succeed in obtaining definitive evidence to support this proposal [11,12]. To address this issue we used the Bacteroides thetaiotaomicron enzyme (anSMEbt), which proved to be more stable and produced three mutants in which groups of conserved cysteinyl residues were mutated to alanyl residues. The following mutants were generated: C24A/C28A/C31A (named mutant M1), C276A/C282A (named mutant M2) and C339A/C342A/C348A (named mutant M3). Mutants were purified, as previously described, starting from a 15 L culture . Purity of the mutants M1 and M2 proved to be satisfactory whereas during the purification of mutant M3, major contamination occurred, probably as a result of proteolytic cleavage (Fig. S3). All purified enzymes exhibited the typical brownish color of [4Fe-4S] 2+ cluster-containing enzymes and a broad shoulder centered near 400 nm (Fig. 3B).
The iron–sulfur cluster content of as-purified and reconstituted samples of WT and M1 mutant anSMEbt were assessed using iron and protein analyses coupled with UV-visible absorption studies of oxidized and dithionite-reduced samples (Fig. S4) and EPR studies of dithionite-reduced samples in the absence or presence of AdoMet (Fig. 4). Samples of as-purified WT and M1 mutant anSMEbt contained 6.3 ± 0.5 and 4.3 ± 0.5 of Fe per monomer, respectively, which increased to 12.0 ± 1.0 and 10.8 ± 1.0 of Fe per monomer, respectively, in reconstituted samples. In all cases the absorption spectra were characteristic of [4Fe-4S]2+ clusters (i.e. broad shoulders centered at ∼ 320 and ∼ 400 nm). Moreover, the extinction coefficients at 400 nm mirror the Fe determinations and indicate 1.6 ± 0.2 and 1.1 ± 0.2 [4Fe-4S]2+ clusters per monomer for the as-purified WT and M1 mutant samples, respectively, and 2.8 ± 0.4 and 2.6 ± 0.4 [4Fe-4S]2+ clusters per monomer for the reconstituted WT and M1 mutant samples, respectively, based on the published range observed for single [4Fe-4S]2+ clusters (ε400 = 14–18 mm−1·cm−1) . The [4Fe-4S]2+ cluster content is likely to be an overestimate for the reconstituted M1 mutant sample as a result of the increased absorption in the 600 nm region, which generally indicates a contribution from adventitiously bound polymeric Fe-S species. While more quantitative analyses will require Mössbauer studies, the analytical and absorption data are consistent with WT and M1 mutant anSMEbt enzymes being able to accommodate up to three and two [4Fe-4S]2+ clusters per monomer, respectively. Hence, the additional seven or eight conserved cysteinyl residues (see Fig. 3A) have the ability to coordinate two additional clusters. A similar conclusion was recently published for the homologous K. pneumoniae AtsB protein based on definitive analytical and Mössbauer studies .
Based on the absorption decrease at 400 nm on reduction, compared with well-characterized [4Fe-4S]2+,+ clusters, we estimate that ∼ 20% and ∼ 30% of the [4Fe-4S] clusters are reduced by dithionite in the reconstituted WT and M1 mutant forms of anSMEbt, respectively (see Fig. S4). Both samples exhibited weak, fast-relaxing EPR signals in the S =1/2 region, accounting for 0.12 spins per monomer for the WT anSMEbt and 0.07 spins per monomer for the M1 anSMEbt (Fig. 4). The relaxation behavior (observable without relaxation broadening only below 30 K) is characteristic of [4Fe-4S]+ clusters rather than of [2Fe-2S]+ clusters. The origin of the low-spin S =1/2 quantifications for dithionite-reduced WT and M1 mutant anSMEbt, relative to the extent of reduction estimated based on absorption studies, is unclear at present. Probably, it is a consequence of [4Fe-4S]+ clusters with S =1/2 and 3/2 spin state heterogeneity as dithionite-reduced reconstituted samples of WT anSMEcpe with substoichiometric cluster content (∼ 6 Fe per monomer) exhibit weak features in the g =4–6 region, indicative of the low-field components of the broad resonances spanning ∼ 400 mT that are associated with S =3/2 [4Fe-4S]+ clusters . As shown in Fig. S5, WT anSMEcpe exhibits well-resolved low-field S =3/2 resonances in the g =4–6 region that are perturbed in the presence of AdoMet, suggesting that the radical-AdoMet [4Fe-4S]+ cluster contributes, at least in part, to the S =3/2 EPR signal. In contrast, the fully reconstituted WT and M1 mutant anSMEbt samples do not exhibit well-resolved resonances in the g =4–6 region (data not shown). However, as indicated below, the lack of clearly observable S =3/2 [4Fe-4S]+ cluster resonances may well be a consequence of broadening as a result of the intercluster spin–spin interaction involving the strongly paramagnetic S =3/2 clusters in cluster-replete samples of reduced anSMEbt.
The S =1/2 resonance for the reduced M1 mutant cannot be simulated as a single species and arises either from two distinct magnetically isolated [4Fe-4S]+ clusters with approximately axial g tensors, or because of a weak magnetic interaction between two [4Fe-4S]+ clusters. We suspect the latter, as two S =1/2 resonances with different relaxation properties cannot be resolved based on temperature-dependence and power-dependence studies. Such magnetic interactions would be expected to be greatly enhanced for clusters with S =3/2 ground states, resulting in additional broadening that would render the resonances unobservable except at inaccessibly high enzyme concentrations. However, irrespective of the explanation of the origin for the complex EPR signal exhibited by the dithionite-reduced M1 mutant anSMEbt, the EPR data support the presence of two [4Fe-4S]2+,+ clusters in addition to the radical-AdoMet [4Fe-4S]2+,+ cluster in anSMEbt. Moreover, subtraction of the reduced M1-mutant EPR spectrum from the reduced WT spectrum affords an axial resonance –g|| = 2.04 and g⊥ = 1.92 – that is readily simulated as a magnetically isolated S =1/2 [4Fe-4S]+ cluster (accounting for 0.05 spins per monomer) and is attributed to the reduced radical-AdoMet [4Fe-4S]+ cluster. This is confirmed by changes in the g values (g =1.98, 1.90, 1.84) and increased spin quantification (0.05 to 0.15 spins per monomer) for the S =1/2 form of the radical-AdoMet [4Fe-4S]+ cluster upon the addition of excess AdoMet (Fig. 4B). Similar changes in the EPR properties of radical-AdoMet S =1/2 [4Fe-4S]+ clusters upon binding AdoMet have been reported for many radical-AdoMet enzymes [28,29], and the increase in spin quantification is likely to be a consequence of the increase in redox potential that results from AdoMet binding . In contrast, within the limits of experimental error, the EPR spectra and spin quantification of the two additional S =1/2 [4Fe-4S]+ clusters that are present in the reduced M1 mutant are not significantly perturbed by AdoMet.
Overall, the EPR and absorption results are best interpreted in terms of three [4Fe-4S]2+,+ clusters in anSMEbt. Each is likely to be mixed spin (S =1/2 and S =3/2) in the reduced state and only one is capable of binding AdoMet at the unique Fe site. As each is only partially reduced by dithionite at pH 7.5, their midpoint potentials are all likely to be in the range of −400 to −450 mV.
Function of anSMEs cysteinyl clusters
Dierks and co-workers carried out pioneering studies to assess the function of the cysteinyl clusters of the anSMEs . They made single amino acid substitutions into the three conserved cysteinyl clusters of anSMEkp and co-expressed the corresponding mutants in Escherichia coli, along with the sulfatase from K. pneumonia. All mutants failed to mature the co-expressed sulfatase as no sulfatase activity could be measured. Nevertheless, it was not possible to conclude whether the mutated enzymes were unable to catalyze any reaction or whether they led to the formation of reaction intermediates such as in spore photoproduct lyase, another radical AdoMet enzyme for which it has been elegantly demonstrated that a cysteinyl mutant, while inactive in vivo , efficiently catalyzes in vitro AdoMet cleavage with substrate H-atom abstraction, leading to the formation of a reaction by-product .
We thus assayed the in vitro activity of WT anSMEbt and mutants after reconstitution in the presence of iron and sulfide. All proteins exhibited UV-visible spectra compatible with the presence of [4Fe-4S] centers (Fig. 3B). Enzymatic assays were conducted using 17C peptide as a substrate and reactions were analyzed using HPLC and MALDI-TOF MS. The results demonstrate that WT anSMEbt is able to mature the substrate peptide, but that none of the mutant forms (i.e. M1, M2, or M3) were able to catalyze peptide maturation or to produce a peptidyl intermediate, as no other peptide was observed by HPLC or MALDI-TOF MS analysis (Fig. 5A,B). Even after derivatization with DNPH, which strongly enhances the signal of the FGly-containing peptide, no trace of modified peptide could be detected using the M1, M2, or M3 mutants (Fig. S6).
AdoMet cleavage was assessed for WT anSMEbt and for the M1, M2, and M3 variants of anSMEbt using the HPLC assay. As expected, the results showed that mutant M1, which lacks the radical AdoMet cysteinyl cluster, is unable to produce 5′-dA, in contrast to the WT enzyme (Fig. 5C). More surprisingly, HPLC analyses revealed that the reductive cleavage of AdoMet was also strongly inhibited in the M2 and M3 mutants, with a 50- to 100-fold decrease observed compared with the WT enzyme (Fig. 5D).
The variant proteins were also incubated with AdoMet under reducing conditions in the absence of substrate, as we previously reported that anSMEbt is able to produce 5′-dA efficiently under these conditions . In the absence of substrate, the AdoMet reductive cleavage activity of all mutants was identical to that obtained in the presence of peptide, again indicating that all three clusters are required for effective reductive cleavage of AdoMet. This observation is most readily interpreted in terms of a role for the two additional [4Fe-4S]2+,+ clusters in mediating electron transfer to the radical-AdoMet [4Fe-4S]2+,+ cluster. A similar interpretation was made to explain the strong inhibition of AdoMet reductive cleavage that was observed in the 4-hydroxyphenylacetate decarboxylase activating enzyme, a radical AdoMet enzyme possessing three [4Fe-4S] centers, when cysteinyl residues in its two additional cysteinyl clusters were mutated to alanines . However, in the absence of detailed spectroscopic characterization of the clusters in the M2 and M3 mutant anSMEbt samples, we cannot rule out the possibility that the loss of one of the additional [4Fe-4S] clusters affects the ability to reductively cleave AdoMet by perturbing the redox potential, AdoMet-binding ability or assembly of the radical-AdoMet [4Fe-4S]2+,+ cluster.
Sequence comparison with other radical AdoMet enzymes
Primary sequence comparisons with previously studied radical AdoMet enzymes did not reveal significant homologies, but several other radical AdoMet enzymes catalyzing post-translational protein modifications contain conserved cysteinyl clusters involved in the coordination of additional [4Fe-4S] centers. These enzymes are B12-independent glycyl radical-activating enzymes (i.e. benzylsuccinate synthase , glycerol dehydratase [34,35] and 4-hydroxyphenylacetate decarboxylase  activases), which catalyze the formation of a glycyl radical on their respective cognate enzyme using 5′-deoxyadenosyl radical. The role of these additional clusters has still to be established, but preliminary mutagenesis studies for a hydroxyphenylacetate decarboxylase activating enzyme indicated a role in mediating electron transfer to the radical-AdoMet [4Fe-4S] cluster .
Further examination of radical AdoMet enzymes involved in protein or peptide modification led to the identification of several proteins sharing the third cysteinyl cluster, Cx2Cx5Cx3C, located in their C-terminal parts while the second cysteinyl cluster found in anSME could only be tentatively assigned in the central part of these proteins (Fig. 6). These proteins are the activating enzyme involved in quinohemoprotein amine dehydrogenase biosynthesis, which is involved in the cross-linking of cysteinyl residues with glutamate or aspartate residues , and a new radical AdoMet enzyme involved in the biosynthesis of a cyclic peptide through a lysine–tryptophan linkage (ST protein) . Although not strictly conserved, we also identified this cluster in PqqE, an enzyme involved in pyrroloquinoline quinone biosynthesis and proposed to catalyze the linkage of glutamate and tyrosine moieties . All these proteins, despite not being homologous, have conserved cysteinyl clusters and catalyze various amino acid modifications. It is thus likely that all these enzymes share common features with anSMEs, and notably the presence of additional [4Fe-4S] centers, as demonstrated for PqqE .
We recently demonstrated that sulfatase maturation catalyzed by the radical AdoMet enzyme, anSME, is initiated by Cβ H-atom abstraction . Nevertheless, the entire mechanism of this enzyme has not yet been deciphered. The results presented herein, using a new anSME substrate, facilitate more definitive conclusions concerning the catalytic mechanism of anSME and the AdoMet requirement. Indeed, using an HPLC-based quantitative assay, we have demonstrated tight 1 : 1 coupling between AdoMet cleavage and FGly production using both cysteinyl-containing and seryl-containing peptides. We also demonstrate the tight inhibition of AdoMet reductive cleavage when the target residue is substituted by an alanyl residue, in contrast to what occurs in the absence of the substrate. Our interpretation is that the peptide binding at the enzyme active site prevents the access of AdoMet to the active site. The recently solved crystal structure of another radical AdoMet enzyme, pyruvate formate-lysase activating enzyme (PFL-AE) , has demonstrated that such a hypothesis is structurally valid. In PFL-AE, the [4Fe-4S] cluster and AdoMet are deeply buried, thereby preventing uncoupling between AdoMet cleavage and glycyl radical generation.
A longstanding question regarding anSMEs concerns the function of the conserved additional cysteinyl clusters originally identified by Schrimer & Kolter . In this bioinformatics study, it was suggested that these clusters were involved in [Fe-S] center co-ordination. The mutagenesis of these conserved residues in the K. pneumoniae enzyme subsequently revealed that they are essential for in vivo activity . Nevertheless, their function remained elusive. Grove et al.  provided the first definitive evidence that they are involved with coordinating two [4Fe-4S] centers in addition to the radical AdoMet [4Fe-4S] center. Based on the inferred AdoMet requirement, a mechanism was proposed involving site-specific ligation of one of the additional [4Fe-4S]2+ centers to the target cysteinyl or seryl residue, resulting in substrate deprotonation. The 5′-deoxyadenosyl radical generated by the reductive cleavage of AdoMet bound at the unique site of the radical AdoMet [4Fe-4S]2+,+ cluster would then abstract a Cβ H-atom from the target residue and an aldehyde product would be generated by using the cluster as the conduit for the removal of the second electron . The proposed mechanism is reminiscent of the isopenicillin N synthase (IPNS), which catalyzes the Cβ-H cleavage from a cysteinyl residue after its co-ordination by a mononuclear nonheme iron center. Following H-atom abstraction, a postulated thioaldehyde intermediate is formed, leading to peptide cyclization [42,43]. Interestingly, using substrate analogs it has been reported that IPNS can oxidize its target cysteinyl residue into a hydrated aldehyde, which is virtually the same as the reaction catalyzed by anSME .
Thus, it is conceivable that one of the two additional clusters binds and deprotonates the target cysteinyl or seryl residues and provides a conduit for removal of the second electron . If such mechanism is correct, our recent demonstration that the 5′-deoxyadenosyl radical produced by anSME directly abstracts one of the cysteinyl Cβ hydrogen atoms , coupled with the results reported herein, indicate that deprotonation occurs before, or simultaneously with, AdoMet cleavage. Indeed, using an alanyl-containing peptide we observed complete inhibition of AdoMet cleavage.
Although the mutagenesis studies reported herein suggest that both of the two additional [4Fe-4S] clusters are required for AdoMet cleavage using dithionite as an electron donor, we cannot rule out the possibility that this is a consequence of perturbation of the redox or AdoMet-binding properties of the radical-AdoMet [4Fe-4S]2+,+ center that are induced by the loss of either of the two additional clusters. Hence, it is possible that one of the additional [4Fe-4S] clusters (Cluster II) is involved with binding the peptide substrate and providing a conduit for removal of the second electron. The other [4Fe-4S] cluster (Cluster III) could function in mediating electron transfer from the physiological electron donor to the radical-AdoMet [4Fe-4S] cluster, or from Cluster II to the physiological electron acceptor, see Fig. 7A. The former mechanism is analogous to that recently proposed by Grove et al. .
Nevertheless, the data presented herein suggest an alternative mechanism. Indeed, the primary sequence analyses discussed above indicate that the two additional clusters are likely to be ligated by the eight conserved cysteinyl residues and hence both [4Fe-4S] clusters may have complete cysteinyl ligation, one cysteinyl residue from the last motif being involved in the co-ordination of the second cluster (Fig. 3A). Furthermore, the preliminary observation that these clusters are almost isopotential with the radical-AdoMet cluster, together with the mutagenesis studies reported herein (which indicate that both additional [4Fe-4S] clusters are required for productive reductive cleavage of AdoMet), suggest that the additional [4Fe-4S]2+,+ clusters play a role in facilitating electron transfer to the radical-AdoMet cluster, as appears to be the case in some B12-independent glycyl radical-activating enzymes . Finally, sequence analysis revealed that these cysteinyl clusters are also found in other radical AdoMet enzymes involved in protein or peptide modification. These enzymes catalyze the modification of amino acids such as glutamate or tyrosine, which are not known to bind [Fe-S] centers. Moreover, another radical AdoMet enzyme, BtrN, has recently been demonstrated to use AdoMet stoichiometrically to catalyze the two-electron oxidation of a hydroxyl group to a ketone without additional Fe-S centers, a reaction formally analogous to the one catalyzed by anSME . However, the absence of additional Fe-S clusters in BtrN clearly requires confirmation using Mössbauer spectroscopy.
Based on the above considerations, we propose an alternate mechanism for anSME (Fig. 7B). In our proposed mechanism, the initial step is the reduction of the radical-AdoMet [4Fe-4S]2+ cluster via electron transfer from the two additional [4Fe-4S]2+,+ clusters. Following this reduction, the Cβ H-atom of the substrate is abstracted by the 5′-deoxyadenosyl radical generated by the reductive cleavage of AdoMet bound at the radical-AdoMet [4Fe-4S]2+,+ cluster, as recently demonstrated . Simultaneously, deprotonation of the thiol or hydroxyl group occurs, catalyzed by an amino acid side chain. The substrate radical intermediate formed by Cβ H-atom abstraction is then further oxidized to yield an aldehyde or a thioaldehyde. In this scenario, the radical would be transferred back to the radical-AdoMet [4Fe-4S]2+ cluster by outer-sphere electron transfer. The implication is that the reaction would have a substrate radical intermediate, as recently demonstrated for BtrN , and would be self-sustaining once the initial electron has been supplied by an exogenous electron donor. Both possibilities are currently under investigation in our laboratories.
For Cys-type sulfatases, both mechanisms shown in Fig. 7 result in the formation of a thioaldehyde intermediate, as is also the case in IPNS  and cysteine decarboxylases [46,47]. Hydrolysis of the thioaldehyde by a water molecule was probably the next step. In accordance with this hypothesis the incorporation of 18O into FGly is observed when the reaction was carried out in H218O buffer (see Fig. S7).
Although further work needs to be carried out to clarify the catalytic mechanism of anSMEs and the role of the two additional [4Fe-4S] clusters, the present report suggests that anSMEs possess common features with some glycyl radical-activating enzymes and that radical AdoMet enzymes possessing additional [4Fe-4S] clusters are likely to be found, notably in enzymes catalyzing protein post-translational modifications. It remains to be seen if the function of these additional clusters involves mediating electron transfer and/or binding and activating the peptidyl substrates.
All chemicals and reagents were obtained from commercial sources and were of analytical grade. AdoMet was synthesized enzymatically and purified as described previously .
anSMEcpe and anSMEbt protein expression and purification
Protein expression and purification were performed as previously described . Briefly, E. coli BL21 (DE3) transformed with a plasmid bearing the anSMEcpe gene or the anSMEbt gene (pET-6His-anSMEcpe or pET-6His-anSMEbt, respectively) were grown aerobically overnight at 37 °C in Luria–Bertani (LB) medium (100 mL) supplemented with kanamycin (50 μg·mL−1). An overnight culture was then used to inoculate fresh LB medium (15 L) supplemented with the same antibiotic. After overnight growth at 25 °C in the presence of isopropyl thio-β-d-galactoside (IPTG), cells were collected and suspended in Tris-buffer (50 mm Tris, 150 mm KCl, 10% glycerol, pH 7.5). The cells were then disrupted by sonication and centrifuged at 220 000 × g at 4 °C for 1 h. The solution was then loaded onto a Ni–nitrilotriacetic acid Sepharose column equilibrated with Tris-buffer, pH 7.5. The column was washed extensively with the same buffer. Some of the adsorbed proteins were eluted by a washing step with 25 and 100 mm imidazole and the over-expressed protein was eluted by applying 500 mm imidazole. Imidazole was removed by gel-filtration chromatography on PD-10 columns (GE Healthcare) and fractions containing the anSMEcpe or anSMEbt proteins were immediately concentrated using Ultrafree cells (Millipore) with a molecular cut-off of 10 kDa.
Construction of cysteinyl cluster mutants
anSMEbt mutants were obtained using the QuikChange site-directed mutagenesis kit (Stratagene). For each mutant a two-step PCR method was used . The following primers were used: for the C24A/C28A/C31A mutant, 5′-GCC GTA GCC AAC CTC GCA GCC GAA TAC GCC TAT TAT-3′ and 5′- ATA ATA GGC GTA TTC GGC TGC GAG GTT GGC TAC GGC-3′; for the C276A/C282A mutant, 5′-GGC GTA GCT ACA ATG GCG AAG CAT GCC GGA CAT-3′ and 5′-ATG TCC GGC ATG CTT CGC CAT TGT AGC TAC GCC-3′; and for the C339A/C342A/C348A mutant, 5′- ACC CAA GCC AAG GAG GCC GAC TTT CTA TTT GCC GCC AAC GGA-3′ and 5′-TCC GTT GGC GGC AAA TAG AAA GTC GGC CTC CTT GGC TTG GGT-3′ (the altered codons are shown in bold). After verification of the correct mutation by sequencing, the plasmids obtained were transformed into E. coli BL21 (DE3) and the mutated proteins were produced using the same protocol as for the WT enzyme.
Reconstitution of Fe-S clusters on anSMEbt and anSMEcpe
Reconstitution was carried out anaerobically in a glove box (Bactron IV). Anaerobically purified anSMEs (200 μm monomer) were treated with 5 mm dithiothreitol (Sigma, St Louis, MO, USA) and incubated overnight with a 10-fold molar excess of both Na2S (Sigma) and (NH4)2Fe(SO4)2 (Sigma) at 12 °C. The protein was desalted using a Sephadex G25 column (GE Healthcare, WI, USA) and the colored fractions were concentrated on Amicon Ultra-4 (Millipore, Billerica, MA, USA). Protein concentrations were determined using the Bradford protein assay (Sigma), with BSA as a standard. Iron concentrations were determined colorimetrically using bathophenanthroline (Sigma) under reducing conditions, after digestion of the protein in 0.8% KMnO4/0.2 m HCl.
The 17-mer peptides (with the critical residue shown in bold) Ac-TAVPSCIPSRASILTGM-NH2, Ac-TAVPSSIPSRASILTGM-NH2 and Ac-TAVPSAIPSRASILTGM-NH2 were synthesized (0.1-mmol scale) using solid-phase methodology on a Rink amide 4-methylbenzhydrylamine resin (VWR, Fontenay-sous-Bois, France) on a 433A Applied Biosystems peptide synthesizer (Applera, Courtaboeuf, France) and the standard Fmoc procedure of the manufacturer. The synthetic peptides were purified by RP-HPLC on a 2.2 × 25-cm Vydac 218TP1022 C18 column (Alltech, Templemars, France) using a linear gradient (10-50% over 45 min) of acetonitrile/trifluoroacetic acid (99.9 : 0.1, v/v) at a flow rate of 10 mL·min−1. Analytical HPLC, performed on a 0.46 × 25-cm Vydac 218TP54 C18 column (Alltech), showed that the purity of the peptides was > 99.1%. The purified peptides were characterized by MALDI-TOF MS on a Voyager DE PRO (Applera, France) in the reflector mode with α-cyano-4-hydroxycinnamic acid as a matrix.
Samples containing 6 mm dithiothreitol, 3 mm sodium dithionite, 500 μm peptides and 1 mm AdoMet, in Tris-buffer, pH 7.5, were incubated with reconstituted proteins. The reactions were performed anaerobically in a glovebox (Bactron IV Shellab, Cornelius, OR, USA). The oxygen concentration was monitored using a gas analyzer (Coy Laboratory, Grass Lake, MI, USA). After incubation at 25 °C, the samples were divided in half: one half was used to test the maturation activity using MS and the other half was used to quantify the reductive cleavage of AdoMet and FGly formation. Control samples were prepared without enzyme to verify peptide and AdoMet stability over time. Experiments performed in H218O were carried out exactly as described above except that the Tris-buffer was made in H218O and the enzyme was exchanged twice with this buffer before the experiments.
Peptide maturation analysis using MALDI-TOF MS
The α-cyano-4-hydroxycinnamic acid matrix (Sigma) was prepared at 4 mg·mL−1 in 0.15% trifluoroacetic acid, 50% acetonitrile. The DNPH matrix was prepared at 100 mg·mL−1 in 0.15% trifluoroacetic acid, 50% acetonitrile. Equal volumes (1 μL) of matrix and sample were spotted onto the MALDI-TOF target plate. MALDI-TOF analysis was then performed on a Voyager DE STR Instrument (Applied Biosystems, Framingham, CA). Spectra were acquired in the reflector mode with 20 kV accelerating voltage, 62% grid voltage and a 120 ns delay.
Peptide maturation and 5′-deoxyadenosine production quantification by HPLC
Peptide modification and 5′-deoxyadenosine production were measured by HPLC using a C18 column (LicroSphere, 5-μm, 4.6 × 150-mm) equilibrated in solvent A (0.1% trifluoroacetic acid). A linear gradient from 0 to 80% acetonitrile was applied at a constant flow rate of 1 mL·min−1. Detection was carried out at 260 nm for AdoMet and its derivative and at 215 nm to follow peptide modification.
X-band EPR spectra were recorded on a Bruker Instruments ESP 300D spectrometer equipped with an Oxford Instruments ESR 900 flow cryostat (4.2–300 K). Spectra were quantified under nonsaturating conditions by double integration against a 1 mm CuEDTA standard.
This work was supported by grants from Agence Nationale de la Recherche (Grant ANR-08-BLAN-0224-02) and the NIH to M.K.J. (GM62524). Mass spectrometry experiments were performed at PAPSSO, INRA, Jouy-en-Josas.